SVMDECON {ADAPTS}

Support vector machine deconvolution

Description

Use SVMDECONV to estimate the cell count percentage David L Gibbs, [e-mail] June 9, 2017

v-SVR is applied with a linear kernel to solve for f, and the best result from three values of v = 0.25, 0.5, 0.75 is saved, where ‘best’ is defined as the lowest root mean squared error between m and the deconvolution result, f x B.

Our current implementation executes v-SVR using the ‘svm’ function in the R package, ‘e1071’.

w2 <- SVMDECON(m, B)

Usage

SVMDECON(m, B)

Arguments

m

a matrix represenging the mixture (genes X 1 sample)

B

a matrix representing the references (genes X cells), m should be subset to match B

Value

A matrix with cell type estimates for each samples